Most AI coding tools collapse the whole software team into a single chat window. This takes the opposite bet: it splits the work across roughly a dozen role-specialized agents — analyst, product manager, architect, developer, QA, scrum master — that hand artifacts to one another the way a real agile team does. The consequence worth noticing is that planning is forced to happen first: a feature becomes a written PRD and architecture document before any code exists, so the development cycle has something concrete to execute against instead of re-deriving intent from a vague prompt each turn.
How It Works
- Two-phase loop: a planning phase produces PRD plus architecture docs, then a development cycle executes against those artifacts — meaning context is captured once in files rather than reconstructed every prompt.
- It runs as a CLI layered onto tools you already use (Claude Code, Cursor, Codex CLI, Gemini) instead of being yet another IDE, so your editor and model choice stay yours.
- A three-layer validation system plus a "squad" mechanism lets you install domain-specific agent packs on top of the core, so the framework extends without forking it.
Who It's For
Great fit if you already work inside an agentic CLI and want structure — explicit roles, documented plans, repeatable handoffs — rather than one improvising assistant. Look elsewhere if you want inline autocomplete or a single-prompt "build me an app" tool: the agile ceremony, the dozen agents, and the document-first flow are real overhead for small one-off changes. Note that the project and much of its documentation are Portuguese-first (Brazilian team), and a paid AIOX Pro tier gates some enterprise features.
